We sell AI chatbots. So we run one. The chatbot you can open in the bottom-right of this site is ours — built on Node.js, Express, and MiniSearch. Zero Claude API cost. Zero hallucination risk. Here is the full stack and the tradeoffs.
Open the live widget on the bottom-right of any page on this site. Ask it about pricing, services, or how to book.
The chatbot embedded on this site is ours — same stack we sell as our Starter tier to clients.
The quickest way to lose credibility selling AI work is to sell what you have never run yourself. So before we sold AI chatbots to a single client, we built one for our own site. The chatbot widget on the bottom-right of every page is the result.
The decision we wrestled with on day one was the same one every small business faces: do we build on a large language model API like Claude or OpenAI, or do we build a lexical-search chatbot with no AI API at all? Both are real "AI chatbots" in the way the market uses the term. They have very different cost profiles and very different tradeoffs.
We chose v1 with no AI API. The reason was simple: most questions a visitor asks on a service business website are predictable. "What do you do? How much does it cost? Where are you based? Can I book a call?" A curated answer file plus a fast lexical search handles 90% of that without ever needing a language model. The cost to run it is $0 per month. The risk of the bot saying something we did not approve is zero, because the bot can only return answers we wrote ourselves.
This site is also where we test the embed flow, the widget UX, and the handoff to contact. Every fix we make on our own chatbot rolls into the Starter tier we sell to clients.
Most small business website chatbots only need to answer a fixed set of questions. A curated answer file plus a fast lexical search handles that for $0 per month and never hallucinates. When the business needs real conversation, we move it to our Pro tier with the Claude API.
Every detail below is pulled directly from the chatbot's package.json and README. Nothing fabricated.
| Runtime | Node.js 20+ |
| Web server | Express.js 4.21 |
| Security headers | Helmet 8 |
| Search engine | MiniSearch 7 — lexical, in-memory |
| Knowledge source | Curated answers.json — approved facts only |
| Hosting | Hostinger Node.js app at chat.ishaanexdigital.com |
| Embed | One <script> tag in the main site Footer |
| AI API cost | $0 / month — no Claude, no OpenAI in v1 |
The chatbot on this site is the Starter tier. The Pro tier adds the Claude API for conversational depth.
Stack: Same stack as the one on this site — Node.js + Express + MiniSearch + curated JSON
Best for: FAQ businesses — clinics, trades, dental, contractors, dog daycare, sanitation. Predictable questions, predictable answers.
Stack: Claude API + retrieval over your real business content + safety guardrails
Best for: Conversational sales, complex services, multilingual conversation, lead qualification flows.
Both tiers can be embedded on any platform — custom-coded sites, WordPress, Wix, or Squarespace. Both tiers ship with the same one-line script tag. The only differences are the backend behind that script and whether a Claude API account is needed. For the full pricing comparison (and the decision tree for which stack fits your business) read the AI chatbot pricing guide. Local Abbotsford businesses can scope and book through the Abbotsford AI automation page.
Try it before you buy it. If the Starter tier handles what your business needs, we ship the same stack on your domain. If you need more, we build the Pro tier with the Claude API.
